Important aspects for the design of efficient high power DFB-BA lasers

1.2.6 Important aspects for the design of efficient high

1.2 Distributed feedback in semiconductor lasers 65 profile with low intensity, for example in the waveguide close to a cladding layer. The introduction of additional impurities and crystal defects due to the fabrication of the DFB grating must be prevented. Finally, radiation loss from higher-order DFB gratings [Str76] [Hen85], [Kaz85] has to be mini-mized by optimizing the duty cycle [Wen04], [Wen06]. To achieve high power conversion efficiency, comparable to a efficiency optimized FP-BA laser, the internal loss of a DFB-BA laser should be the same as for the FP-BA laser.

High internal quantum efficiency

Basically, an epitaxy design should be optimized for a high internal quantum efficiency. This requires sufficiently high barriers to the quantum wells in the active region, in order to minimize current leakage effects. Furthermore, the DFB grating should be optimized to prevent carrier recombination which would reduce the stimulated emission in the quantum well active region. To achieve high power conversion efficiency, comparable to a efficiency optimized FP-BA laser, the internal quantum efficiency of a DFB-BA laser should be the same as for the FP-BA laser.

Low voltage

A low voltage drop over the laser diode must be achieved at high power op-eration currents for a high efficiency. The usage of semiconductor materials with a low band-gap in the waveguide structure and a sufficient doping pro-file helps to reduce the series voltage and series resistance. The resistivity of the epitaxial layer system also depends on its thickness. Clearly, there is a tradeoff between the aim to achieve a ow internal optical loss (requires a low doping profile) and the aim to achieve a low voltage (requires a high doping profile). Additional semiconductor layers and material interfaces should be reduced as much as possible in order to achieve a low series voltage and series resistance. Finally, the n-substrate, p-contact layer and the ohmic contacts to the n- and p-metallization have to be considered. To achieve high power conversion efficiency, comparable to a efficiency optimized FP-BA laser at the same wavelength, the defect voltageU_def(I) = U(I)−hν/q of a DFB-BA laser should be the same as for the FP-BA laser.

Spectral properties

A CW driven DFB laser is spectrally locked to the Bragg wavelength of the DFB grating over a certain range in injection current and temperature. This requires, that the gain spectrum is spectrally optimized to the Bragg wave-length of the DFB grating. As already discussed in section 1.1.5, the peak of the gain spectrum and the Bragg wavelength should be optimized to co-incide at the required optical output power and at a given temperature for

a high efficiency. This requires that the grating period, the effective index of refraction and the gain spectrum has to be optimized. Because the main aspect of this work is to achieve high power conversion efficiencies, the opti-cal feedback from the DFB grating will be chosen in order to achieve a high slope efficiency, instead of choosing it for a maximum suppression of lasing on FP-like modes. Therefore, the reflectivity of the front facet must be very low and is the key parameter for the suppression of FP-like modes.

Limited operation range

From the discussion of the spectral properties of a CW driven DFB laser it turns out, that its operation will be limited to a specific range in tem-perature and current, in order to achieve the required high optical output power, high power conversion efficiency, wavelength stabilization and narrow spectral width. This range in temperature and current presumably becomes smaller, if the coupling strength is decreased and broader, if the front facet reflectivity is decreased.

To conclude, the development of high power and high efficiency DFB-BA lasers can be done by approximately adapting the properties of the DFB-BA laser to the properties of a power and efficiency-optimized FP-DFB-BA laser.

Firstly, the DFB resonator loss should be comparable or slightly smaller than the FP mirror lossα_DFB ≤α_m. Secondly, the internal loss and internal quantum efficiency and transparency current density should be comparable.

As derived above, this will lead to the same threshold current and slope efficiency for the DFB-BA and FP-BA laser. Based on this requirements, comparable power conversion efficiency will be obtained, if the voltage drop U(I) is approximately identical. Finally, the wavelength of the gain-peak must be optimized to coincide with the Bragg wavelength of the DFB grat-ing at the desired output-power. The qualifications, specified above, imply high requirements to the laser and grating design, fabrication technology and material quality.

Chapter 2

DFB-BA laser review

In this chapter, physical properties and technology of previously published work on typically high power DFB-BA lasers before the beginning of this work are pre-sented. These lasers were usually developed to achieve a narrow spectral width, highly suppression of side modes in the optical spectrum and wavelength stabi-lization over a wide range in temperature. Additionally, the achievable optical output power was higher than what could be obtained from a DFB-RW laser in transversal and longitudinal single mode emission. The possibility to achieve high optical output power ≥4 W and power conversion efficiency ≥50 % from DFB-BA lasers has not been a major matter of research and development, since first promising results in this field have been reported by Kanskar et al. [Kan06].

Good understanding of the best previously published DFB-BA lasers, their design and fabrication technology is essential for further successful development. This chapter summarizes which power and efficiency levels were previously achievable and reviews promising fabrication technologies for DFB gratings to be used in wavelength stabilized BA lasers. Thus, findings from the review concerning the efficiency, optical output power and wavelength stabilization can be used as as a design guidance in this work.

After distributed feedback has been initiated as a principle for optical feed-back in a gelatine film dye laser by Kogelnik et al. [Kog71] in 1971, this has been rapidly transferred to semiconductor diode lasers [Nak73], [Sci74].

Early DFB diode lasers were typically fabricated using a narrow ridge waveg-uide design in order to obtain transversal single mode emission, in addition to longitudinal single mode behavior [Nak89], [Wen04]. Other fabrication approaches, such as buried or diffused waveguides have been reported as well [Sin93], [Lam98]. Later, DFB-RW lasers have been used and devel-oped for telecommunications technology [Ber02], [Fun04] and spectroscopy

applications [Sum12], for example. These lasers were typically optimized for lateral and longitudinal single mode lasing, a high side mode suppression ratio (SMSR) and spectral stabilization over a wide range in temperature.

Achieving high optical output power was no main focus of the development and typical values in the<1 Wrange have been achieved [Wen04]. Even the possibility to achieve high power conversion efficiency has not been consid-ered very seriously. Such narrow stripe width DFB lasers must be considconsid-ered as a basis for the development of DFB-BA lasers but they are rather not the main aspect of this chapter.

In parallel to the development of narrow stripe DFB lasers for transversal single mode emission, the question arises how the optical output power can be increased to>1 W. Distributed feedback BA lasers were fabricated based on the obvious assumption, that the achievable optical output power can be increased by increasing the width of the laser stripe. This first types of DFB-BA lasers were primarily developed to achieve a narrow spectral width, good suppression of side modes in the optical spectrum (for example > 30 dB SMSR) and wavelength stabilization over a wide range in temperature at optical output powers above 1 W.

The first gain coupled 923 nm DFB-BA lasers were demonstrated by Robadey et al. [Rob97]. These lasers consist of a GaAs waveguide with InGaAs quantum wires and AlGaAs cladding layers and were fabricated, based on a two-step epitaxy process. The molecular beam epitaxy (MBE) growth process is interrupted within the GaAs waveguide and a V-groove second-order grating is etched into it with lithographic techniques. After be-ing returned into the MBE reactor, oxide desorption is applied at ∼600^◦C and the surface is regrown with a thin GaAs layer andInGaAsis grown into the V-grooves to form quantum wires (periodicity of264 nm), which provide periodic gain and function as a second-order DFB gain grating. From this material, DFB-BA lasers with a stripe width up to 100µmwere fabricated.

The authors present spectra with good side mode suppression and a spectral width of 1 nm. From the width of the stopband, they determine a cou-pling coefficient of∼30 cm⁻¹. However, measurements had to be performed at liquid nitrogen temperature of77 Kunder pulsed current injection and val-ues of the optical output power have not been reported. These lasers could not exceed threshold at room temperature (even on FP modes), presumably because of too high optical loss and too low gain, provided by the quantum wires and regrowth technology.

Earles et al. [Ear98] successfully increased the optical output power of DFB-BA lasers at 893 nm above 1 W. The epitaxy design is based on an InGaPwaveguide, anInGaAsDQW andInGaAlPcladding layers. A second-order DFB index grating is etched 50 nm deep into the p-type waveguide

69 with lithographic techniques outside the metal-organic chemical vapor depo-sition (MOCVD) reactor, then overgrown with the InGaP p-type cladding layer. Thus, the authors also apply a two-step epitaxy process, but use the difference of index of refraction between the waveguide and cladding layer material to provide periodic optical feedback, instead of structuring the ac-tive region. Compared to the design and technique used by Robadey et al.

[Rob97], this method has the advantage that the DQW active region can be grown with high quality and is unaffected by the etching and two-step epitaxy process. The grating is etched into aluminum-free material which is reported to prevent problems with the formation of stable aluminum oxide and oxygen contamination at the grating interface. The authors fabricated DFB-BA lasers from the material with 100µmstripe width and 1 mmcavity length. A coating with a front-facet reflectivity of R_f = 5 % and rear-facet reflectivity ofR_r = 95 % has been applied to the cleaved facets. Continuous wave optical output power of ≈ 1.1 W is obtained at ≈ 1.9 A with a power conversion efficiency of 32 %. The authors also report on the optical spec-trum, which consists of several lateral modes that originate from a single longitudinal mode. They propose that lasing in a single longitudinal mode has been obtained due to using κ ≈ 7 cm⁻¹ which leads to a low coupling strength κ·L ≈ 0.7 cm⁻¹. Coupling strength between 0.5 and 1.0 is stated to lead to relatively uniform longitudinal field profiles and therefore avoids longitudinal spectral hole burning. The authors further recognize that low coupling strength and an asymmetric facet coating will lead to high differen-tial quantum efficiency.

Further advancement in the technology of BA diode lasers with DFB index gratings has been reported by Chang et al. [Cha00], where some of the authors from [Ear98] are also involved. The epitaxy design has been changed towards a completely aluminum-free waveguide structure that con-sists ofInGaPcladding layers, anInGaAsPwaveguide and anInGaAsDQW active region. Again, the second-order DFB grating is structured into p -type waveguide region outside the MOCVD reactor, then overgrown with aluminum-free cladding compound. This aluminum-free epitaxy design was developed for > 10 W (CW) high power FP-BA diode lasers as reported in [Maw96] and [AM98] and is originally based on the aluminum-free large op-tical cavity (LOC) design, published in [Bot96]. DFB-BA lasers with100µm stripe width,2 mm cavity length and facet coatings forRf = 5 %, Rr = 95 % achieve optical output power of 1 W with a power conversion efficiency of 33 % and a narrow spectrum (< 1 nm) at 976 nm. The authors determine a coupling coefficient of κ ≈ 1.5 cm⁻¹ and thus, κ·L ≈ 0.3 is quite low.

Finally, the authors also determine the internal optical loss under DFB oper-ation (6 cm⁻¹) and under FP operation (4 cm⁻¹). Even4 cm⁻¹ is higher than

the internal optical loss of the same device structure without the grating, reported in [Maw96] and [AM98]. Presumably, the difference occurs due to significant radiation loss from the second-order DFB grating (non-ideal duty cycle) or increased internal optical loss.

In the publications, mentioned above, DFB-BA diode lasers, wavelength stabilized at three different wavelength have been reported. Robadey et al.

[Rob97] report on 923 nm DFB-BA lasers but the do not explicitly point out possible applications of this wavelength. Earles et al. [Ear98] fabricate 893 nmDFB-BA lasers for polarization of Cs which is used for the generation of spin-polarized Xe gas. Furthermore, Changet al. [Cha00] report on976 nm DFB-BA lasers which could be promising pump sources for Yb-doped solid state lasers or fiber lasers.

Klehr et al. [Kle06] have developed DFB-BA lasers emitting at 808 nm, which can be used for pumping Nd:YAG solid state lasers. Besides introduc-ing a new wavelength for DFB-BA lasers, the authors usedAlxGa1−xAs-based designs with aluminum-free grating layers.

Meanwhile, epitaxy designs with aluminum-free waveguide [Ear98] and even completely aluminum-free designs [Cha00] have been developed and were stated to be favorable for the monolithic integration of the DFB grating with a two-step epitaxy process, because exposing aluminum to air during the grating fabrication would lead to the formation of stable oxides and could cause oxygen contamination in the regrown material. Note, that in contrast to oxides onInGaAsP, aluminum oxides cannot be desorbed by a temperature increase to ∼ 700^◦C in the epitaxy reactor. In parallel to aluminum-free designs for high power FP-BA lasers,(In)AlGaAs-based epitaxy designs have been developed, showing promising results in power and efficiency [Sak92b], [Sak92a]. Advantages and disadvantages of both material systems are for example discussed in [Tre00].

Klehr et al. have implemented a multiple layer DFB grating into the p-side of the 2000 nm broad Al0.45Ga0.55As waveguide, 600 nm above the single quantum well active region, located in the center of the waveguide.

The wafers were grown using low-pressure metal-organic vapor phase epitaxy (MOVPE). The cladding layers consist ofAl0.70Ga0.30As. The first epitaxy is finished with a aluminum-free layer system fromIn_0.48Ga_0.52P,GaAs_0.85P_0.15 and In_0.48Ga_0.52P. The second-order DFB grating is etched into the upper two layers of this aluminum-free layer system with lithographic techniques.

The three-layer approach allows the usage of selective etching techniques, so that the last In_0.48Ga_0.52P layer can stop etching and uncovering of the underlyingAl0.45Ga0.55Asis avoided. After the ex-situ etching, the surface is regrown with the second part of the p-side waveguide and cladding layer to complete the structure. Thus, the authors combine the benefit of

aluminum-71 free grating material (avoid formation of stable aluminum oxide and oxygen contamination in the regrown compound) with the benefit of Al_xGa1−xAs -based waveguide and cladding layers. Klehr et al. experimentally determine a coupling coefficient of ∼ 1 cm⁻¹. DFB-BA lasers with a stripe width of 100µmand a cavity length of 3 mm from this material have therefore a low coupling strength κ·L ≈ 0.3. Facet coating has been applied to achieve a front-facet reflectivity ofR_f <0.1 % and rear-facet reflectivity ofR_f = 95 %. Single emitter DFB-BA lasers were investigated at 20^◦Cunder CW current injection and found to have threshold current of0.58 Aand a very high slope efficiency of 1.06 W A⁻¹. A peak output power of 4.7 W has been obtained at≈5.8 Awith 34 %power conversion efficiency, until power saturation and roll-over is observed. Wavelength stabilization is reported for the whole in-vestigated current range (0- 6 A) with a spectral width <0.3 nm.

The epitaxy design and grating technology for 808 nm DFB-BA lasers, reported in [Kle06], has later been transferred to 975 nm DFB-BA lasers in [Sch09a]. These lasers were optimized to achieve a narrow vertical far field and for wavelength stabilization over a wide temperature range. The idea behind the development of narrow, far field DFB-BA lasers was to en-hance the possible coupling efficiency of the laser emission to lenses or optical fibers. For narrow far field emission, an asymmetrical LOC design has been used for the 4.8µm wide Al0.35Ga0.65As waveguide, embedded between n -and p-cladding layers from Al_0.45Ga_0.55As and Al_0.85Ga_0.15As, respectively.

The Al_0.35Ga_0.65As waveguide consists of a 3.9µm thick n-type region and a 0.9µmthick p-type region with an InGaAs DQW active region located in between. The second-order DFB grating (295 nm period, duty cycle opti-mized to c_d ≈ 1/4) is integrated with a two-step epitaxy process into the p-type waveguide,0.8µm above the DQW and 0.2µm under the p-cladding layer. As reported in [Kle06], it consists of a three layer system, here using InGaP/GaAs/InGaP, grown at the end of the first epitaxy. The grating is structured with lithographic techniques and selective etching is applied to stop on the lower InGaP layer. Afterwards, the wafer is transferred back to the MOVPE reactor and the surface is regrown withAl_xGa1−xAsto complete the structure. The coupling coefficient has been experimentally determined as ≈ 9 cm⁻¹. DFB-BA lasers with a stripe width of 100µm and a cav-ity length of 2 mm have high κ·L ≈ 1.8 and were facet coated to achieve Rf <0.1 % andRr = 95 %. Mountedp-side down, these lasers achieve an op-tical output power of2.4 Wat4 ACW and25^◦C. From the voltage and power characteristics, a threshold current of0.34 A, a slope efficiency of0.69 W A⁻¹ and a peak power conversion efficiency of 35 % has been determined. From the length dependence of the differential quantum efficiency and threshold current density, the internal optical loss α_i and internal quantum efficiency

η_i was determined both – for DFB-BA lasers (spectrally detuned to achieve lasing on FP modes) and FP-BA reference devices, grown in a single stage epitaxy without the grating layers. While ηi = 0.85was found for the DFB-BA and FP-DFB-BA lasers,α^DFB_i = 2.0 cm⁻¹ for the DFB-BA lasers is increased, compared to α^FP_i = 1.2 cm⁻¹, found for the FP-BA lasers. The asymmetric LOC waveguide design results in a narrow vertical far field emission angle of 14^◦ FWHM and38.5^◦ with95 % power content, which is significantly lower than what has been reported earlier for other DFB-BA lasers. The high cou-pling strength ofκ·L≈1.8is a compromise which on the one hand, leads to a compromised slope efficiency, but on the other hand, enables wavelength stabilization between 15 and 100^◦C via a low threshold current and good suppression of FP-like modes [Sch09a], [Sch09b].

From about 2006 onwards, DFB-BA lasers began to be optimized pri-marily for optical power (≥ 4 W) and power conversion efficiency (≥ 50 %) to make them suitable for high power pumping applications. This type of DFB-BA lasers was developed to obtain comparable optical output power and power conversion efficiency as already achieved with FP-BA lasers [Sak92b], [Sak92a], [Bot96], [Maw96], [AM98], [Tre00], [Kan05], [Kni05]. In order to maximize the optical output power and power conversion efficiency, some degradations in the temperature range with sufficient wavelength stabiliza-tion and side mode suppression must be accepted.

In 2006, Kanskar et al. [Kan06] published experimental results which have been obtained with 975 nm DFB-BA lasers (suitable for pumping Yb doped gain media). Even though the authors conceal the material system in use, one may assume that the lasers have been developed, based on the aluminum-free epitaxy design they have published in earlier work [Cha00].

The second-order DFB grating is again placed between thep-type waveguide and the p-type cladding layer and structured by use of a two-step epitaxy, lithographic techniques and 100 nm deep reactive ion etching. Cleaved into cavity length of1 and 2 mm, 100µm broad DFB lasers were facet coated to achieve R_f ≈ 4 % and R_r ≈ 95 % and mounted p-side down. The authors note thatκ·L∼1(presumably for1 mmcavity length). DFB-BA lasers with 1 mm cavity length have a slope efficiency of 0.9 W A⁻¹ and achieve ≈ 4 W CW peak power and a peak power conversion efficiency of53 %at2 Wat25^◦C heatsink temperature. Furthermore, 2 mm long DFB-BA lasers were driven up to8 Aand achieve5.5 Woptical output power. Finally, spectra at2 Aare shown for heatsink temperatures of10,20, . . .50^◦Cand provide locking over 40^◦C. The spectra are structured due to the appearance of multiple lateral and maybe even multiple longitudinal modes and the spectral width remains

<0.5 nm.

Furthermore, part of the authors from [Kan06] report on 808 nm

DFB-73 BA lasers in He et al. [He09]. Presumably, the authors have adapted the epitaxy design from [Kan06] to be suitable for 808 nm DFB-BA lasers (now suitable for pumping Nd:YAG). In fact, the authors again conceal the mate-rial system in use. The grating technology is also comparable to what has been published in [Kan06]. DFB-BA lasers (100µm×2 mmwere again facet coated to achieve Rf ≈ 4 % and Rr ≈ 95 % and mounted p-side down, the coupling strength isκ·L∼1. At25^◦C, these lasers have a threshold current of 0.45 A, a slope efficiency of > 1 W A⁻¹ and achieve 4 W optical output power with a power conversion efficiency, close to the peak efficiency of 57 % (at 3 W). The spectral width is shown to be ∼0.5 nm when it is evaluated at approximately95 %power content. The authors finish their report with a diagram, showing sporadic measurements of the spectral locking range over heatsink temperature (between10and50^◦C) and CW drive current (between 0 and 4.5 A). Interestingly, it is limited at high temperatures and currents, as well as for low temperatures. The locking range in heatsink temperature is indeed, lower than for a DFB-BA laser with κ·L > 1, as for example, reported in [Sch09a].

Finally, the question arises what can be learned from the publications about DFB-BA lasers mentioned above, which have been reported before or at the beginning of this work, with respect to the development targets in this work. The most important findings are itemized below.

• The two-step epitaxy is a successful concept for the monolithic integra-tion of DFB gratings, although it requires a sophisticated fabricaintegra-tion technology.

• Usage of second-order DFB gratings is a proved concept for ease of fabrication and the corresponding radiation loss from first-order Bragg scattering in this gratings can be small if the duty cycle is adjusted precisely.

• The epitaxy design of the waveguide structure can be based either on an aluminum-free material system or on an AlGaAs-based material system, both have achieved promising power and efficiency results in FP-BA lasers and laser bars.

• For high power, high efficiency DFB-BA lasers, long cavity length and low coupling strength κ·L≤ 1 is preferred in order to achieve a high differential quantum efficiency.

• The front facets of DFB-BA lasers must be AR coated for a sufficient suppression of lasing on FP-like modes at the gain peak wavelength

which is essential to achieve a narrow wavelength stabilized spectrum in a broad range of injection current and heatsink temperature.

Chapter 3

Experimental and theoretical results from iteration I

In this chapter, conclusions drawn in chapter 1 will be applied to the develop-ment of high power and efficiency DFB-BA lasers. Firstly, a design for FP-BA reference lasers is presented. These lasers have been optimized for high optical output powers and high power conversion efficiency, as well as for the monolithic integration of a DFB grating. Based on the design specifications, lasers were grown, processed, facet coated and mounted, in order to enable extensive device characterization. Voltage and power characteristics were then investigated, as well as the spectral properties and the temperature dependence of the wave-length, threshold current and slope efficiency. The characterization is completed with investigation of the spatial emission properties. The experimental results show, that these lasers are promising for state of the art FP-BA diode lasers and potentially enable the further development of DFB-BA lasers in the∼10 W output power, ∼60 % power conversion efficiency range.

In the second part of this chapter, the development of a DFB grating for efficient high power DFB-BA lasers is described. DFB-BA lasers with such gratings are then fabricated and characterized, comparable to the FP-BA reference lasers.

DFB-BA lasers are investigated in order to analyze similarities and differences, compared to the reference devices. Finally, the power, voltage and efficiency char-acteristics of the best DFB-BA lasers from iteration I are compared to results, obtained with the numerical calculations. Conclusions are drawn, how the power conversion efficiency of the DFB-BA lasers can be further optimized in order to approach to the properties of the highly efficient FP-BA reference lasers.